The now-withdrawn Lewandowsky Fury paper (link) is possibly one of the egregious examples of ethically compromised research encountered. Delve into the paper, the first thing crossing one’s mind is: how did the university ethics committee approve this project? This was the study protocol – Lewandowsky’s associates would carry out real-time surveillance on people criticizing his paper, prod and provoke them, record their responses and perform ‘analysis’. How did they say yes?
Lewandowsky’s correspondence with University of Western Australia (UWA) officials has been released (link). Amidst a storm of emails on this previous work, he writes to the secretary of the ethics committee (10 Sep 2012) of his intention to start another project:
This is just to inform you of the fact that I will be writing a follow-up paper to the one that just caused this enormous stir. The follow-up paper will analyze the response to my first paper …
Lewandowsky states there will be no interaction with his subjects: none of the research “will involve experimentation, surveys, questionnaires or a direct approach of participants of any sort“. (emphasis mine)
What would the research be? According to Lewandowsky, his team would “analyz[e] “Google trends” and other indicators of content that are already in the public domain (e.g. blog posts, newspapers, comments on blogs, that type of thing)”. The research would “basically just summarize and provide a timeline of the public’s response.”
The email is a remarkably misleading and limited description of the project he and his associates conducted.
The ethics office response is further divorced from reality. The approval was granted as a “follow-up” study to the ‘Moon’ paper. The ‘Moon Hoax’ paper was itself was approved under an application for “Understanding Statistical Trends”. As recounted here, “Understanding Statistical Trends” was a study where Lewandowsky’s associates showed a graph to shopping mall visitors and asked questions (link pdf). This application was modified to add the ‘Moon hoax’ questions on the day the original paper was accepted for publication. The same application was modified for the ‘Recursive Fury’ paper. Each modification introduced ethical considerations not present in the previous step. Nevertheless, three unrelated research projects were allowed to be stacked on to a single ethics approval by the university board. In this way, Lewandowsky was able to carry out covert observational activities on members of the general public, as they reacted to his own work, with no human research ethical oversight.
Lewandowsky pitches his study proposal as non-intrusive, observational and retrospective in design: there is “no human participation”, the “content is already in the public domain”, and “irrespective of whether we then summarize that activity”. What he implied was there was minimal concern for more elaborate safeguards and vetting usually put in place when working with human subjects.
Yet during the period of study, Lewandowsky was in direct conversation with his study subjects (even as he ostensibly observed them). On a posting spree, he wrote 9 articles at shapingtomorrowsworld.org between Sept 3 – 19, 2012. About half of these were written after he approached the ethics office on the 10th. All but two were written after he announced that he was already collecting data, to the university deputy vice chancellor on the 5th. Among individuals named in the paper as harboring conspiracist ideas, three posted detailed comments with multiple questions responding to these posts, on his website. The subjects wrote numerous posts at their own blogs on Lewandowsky’s actions in the same interval. The flow of comments, appearance and final content were influenced by the second author, John Cook. A team headed by Cook operated as moderators at shapingtomorrowsworld.org, deleting parts, or whole comments offered by the subjects in the same interval. The elicited comments and posts were harvested as data for the paper.
The study was thus not an examination of archived material on blogs. As the authors themselves describe, they recorded subject comments and blog-posts in “real-time”, responses occurring to events set in motion by themselves. It cannot be considered a observational study either as authors interacted with the purported subjects during the period of study.
In her reply, the ethics secretary directs Lewandowsky to the UWA Human Subjects research web page (link). The page contains a ‘risk assessment checklist’ to guide researchers to whether a planned study would need ethics approval. It has these questions:
- Active concealment of information from participants and/or planned deception of participants
- Will participants be quoted or be identifiable, either directly or indirectly, in reporting of the research?
- Will data that can identify an individual (or be used to re-identify an individual) be obtained from databanks, databases, tissue banks or other similar data sources?
- Might the research procedures cause participants psychological or emotional distress?
- Does the research involve covert observation?
The answer is a ‘Yes’ to many of these questions. ‘Participants’ declared to be conspiratorial by Lewandowsky are directly identified by name in the paper. The element of covert observation is undeniable.
The possibility of ethical breaches with internet-based research are well-understood. Clare Madge (2007) observed ethically questionable research could come to be carried out “under the radar screens of ethics committees” simply owing to the ease and speed of internet-based research resulting in ‘shoddy cowboy research’ and proliferation of ethical misconduct. The study design and conduct of the Lewandowsky et al 2013 ‘Recursive Fury’ contains numerous ethical failures. Lewandowsky’s email characterized his work in terms which turned out to be their opposite. There was no formal application and there was no review and consequently the prospective,non-observational nature of his project went unscrutinized.
Madge C. Developing a geographers’ agenda for online research ethics Prog Hum Geogr Oct 2007; 31(5): 654-674